📣 HyperGAI Unveils Its Own Language Model, HPT
posted 19 Mar 2024
HyperGAI, a company specializing in artificial intelligence, has released a new multimodal language model (Multimodal LLM) named HPT, capable of recognizing images and videos. According to the developers, the HPT Pro version outperforms competitor models such as GPT-4 and Gemini Pro in MMBench and SEED-Image performance tests.
HPT's unique selling point is its compatibility with additional pre-trained LLMs, enhancing its functionality. Yet, HyperGAI has opted to publicly release only the source code for the more basic iteration, HPT Air. This version is theoretically capable of handling a wide range of daily tasks for both individual users and businesses alike.
The capabilities of HPT are broad, ranging from recognizing people and objects in photographs, explaining graphs, to solving math puzzles, and interpreting conceptual art. Additionally, HPT can generate creative content such as recipe ideas, travel advice, or even write narrative stories based on images alone.
It's worth noting that prior to this, the company xAI made headlines by open-sourcing their language model Grok-1, and Inflection AI launched its own creation, Inflection-2.5, for the personal assistant Pi, which was noted for producing responses that closely mimic those generated by the AI, Claude.